About DEMO-EM2
DEMO-EM2 is an improved method of DEMO-EM to automatically construct protein complex structures from cryo-EM density maps through
iteratively assembly procedure intertwining chain-level matching, domain-level matching, and domain-based fitting based on
fast quasi-Newton optimization and Differential Evolution algorithms.
Figure 1. Pipeline of DEMO-EM2 for protein complex structruces modelling
DEMO-EM2 is freely available for both academic and commercial users. If you have any questions about DEMO-EM2, please feel free to contact us at
zxg@zjut.edu.cn
Download DEMO-EM2
The download link below contains the program of DEMO-EM2 and a test case (PDBID: 8PUH, EMD-17943).
Click here to download DEMO-EM2 standalone package
File list for DEMO-EM2:
DEMO-EM2 | : main program of DEMO-EM2 |
DomainOptimize | : program for domain-level optimization |
DomainMatching | : program for domain matching |
ModelMatching | : program for chain matching |
domainSplit | : program for domain splits |
RemoveMapRegion | : program for removing map regions matched by structures |
CalCC | : program for assessing the correlation between density map and model |
example | : file of an example (8PUH) containing individual chain models and density map |
How to Run DEMO-EM2
DEMO-EM2 is an automatic protein complex model building program.
Given a density map and the corresponding protein sequences of individual chains,
The 3D structures of individual chains are predicted from sequences
using external protein structure prediction programs.
Utilizing the density map and the predicted models of individual chains,
DEMO-EM2 automatically assembles the protein complex structure.
In this tutorial, we will walk through a simple example using EMD-17943 (PDBID:8PUH).
Step 1: perparing input chain models for DEMO-EM2
Starting from the target protein sequence, 3D structures of individual chains are generated
by a protein structure prediction program.
We use
AlphaFold2
to build structures for chains in this work. Many other programs, such as
I-TASSER,
trRosetta,
and
RoseTTAfold could also
be employed for chain modeling.
Note:
the tutorial for runing AlphaFold2 can be found in
its original paper.
In our experiments, all templates released before the date of each test case are excluded by
setting the "
max_template_data" in the AlphaFold2 runing script.
However, users should ignore this setting to generate the best model in real applications.
Figure 2. Individual chain structures modelled by AlphaFold2 for the example, named as chain1.pdb and chain2.pdb
Step 2: constructing protein complex models by DEMO-EM2
With the individual chain models and the corresponding map, DEMO-EM2 could automatically build
the complex model of the target protein.
Usage: DEMO-EM2 
data_path 
map_path 
resolution 
[Options]
Required arguments:
data_path  | : folder containing chain structures to be assembled
|
maph  | : input density map
|
resolution  | : resolution of the input density map
|
Options:
-ipn  | : name of the chain structure (default: chain) |
-opn  | : output name of the final protein complex model, format: XXX.pdb (default: Cmodel.pdb) |
-lco  | : flag to use local domain optimization, 0: no, 1: yes (default: 1) |
-glo  | : flag to use global domain optimization, 0: no, 1: yes (default: 1) |
-dmm  | : flag to use domain matching, 0: no, 1: yes (default: 1) |
-nlb  | : number of positions to be explored by LBFGS (default: 20) |
-rsc  | : cutoff of correlation score in fraction space (default: 0.6) |
-gw  | : grid width of the map, 0:original value (default: 0) |
-dc  | : threshold of density value after normalization, (default: 0.005) |
-sl  | : flag to output intermediate information, 0: no, 1: yes (default: 0) |
-rm  | : flag to remove matched map regions by models, 0: no, 1: yes (default: 1) |
-it  | : flag to use iterative assembly, 0: no, 1: yes (default: 1) |
Notes:
- All information listed above can be displayed by type
$ ./DEMO-EM2
.
Please do not change the default settings unless you know what you are doing.
- If "-ipn" is set as "chain", models to be assembled should be named as "chain1.pdb, chain2.pdb, chain3.pdb,..." in the data_path.
- "-rsc" should be within the range of (0,1), default: 0.6. It is used to define a good mathcing between model and map.
- "-dc" should be within the range of (0,1), default: 0.005. It is employed to remove unuseful voxels in the map.
- Output models are named as "Cmodel_X.pdb" and saved in the data_path.
- Users can run DEMO-EM2 for multiple times to generate multiple complex models using differnt -rsc cutoff and select a best one as the final model.
Example usages:
$ ./DEMO-EM2 ./eaxmple ./eaxmple/map.mrc 6.2
Output files:
Cmodel1.pdb | : the final complex model constructed by DEMO-EM2 |
chainX_dom.def | : the domain definition of the X-th chain |
Figure 3. Final complex model constructed by DEMO-EM2 and shown within the map
Step 3: Evaluating the quality of the model
Users can calculate Correlation Coefficient (CC)and integrated Fourier Shell
Correlation (iFSC) between the model and density map to mesure the quality of fitness .
Usage: CalCC 
map_path 
model_path 
resolution 
[Options]
Required arguments:
map  | : input density map | |
model  | : DEMO-EM2 constructed model | |
resolution  | : resolution of the density map | |
Options:
-tp  | : atom type for correlation score calculation, "CA", "BB", "all" (default: CA) |
-gw  | : grid width of the map, 0: original value (default: 0) |
-rc  | : types of score, "cc","rscc","fc","ifsc" (default: cc) |
-sl  | : flag to output intermediate information, 0: no 1: yes (default: 0) |
Notes:
- "-tp" supports "CA","BB", "all", where they indicate CA atoms, backbone atoms, and all atoms, respectively.
- "fc" indicates fast density score.
- "cc" indicates correlation coefficient score.
- "rscc" indicates correlation score in fraction space.
- "ifsc" indicates integrated Fourier Shell Correlation.
- All scores are the higher the better.
- The program will interpolate a new map with voxel size equal to the given value if the provided grid width is greater than the original one.
Example usages:
$ ./CalCC ./eaxmple/map.mrc ./eaxmple/Cmodel1.pdb 6.2 -rc ifsc -tp BB
DEMO-EM2 Download
DEMO-EM2 Resource:
References:
-
Ziying Zhang, Yaxian Cai, Biao Zhang, Lydia Freddolino, Guijun Zhang, Xiaogen Zhou.
DEMO-EM2: Assembling protein complex structures from cryo-EM maps through intertwined chain and domain fitting,
submitted, 2023.